simulation multivariate ordinal data with 10 variables

[thanoon]

hi everyone

i have a problem with simulation using sas/iml and i cannot get on data just x1-x10 without data anyone help me please to correct this program.

thanks alot

%include "RandMVOrd.sas";

proc iml;

load module=_all_;                     /* load the modules */

    /* P1    P2     P3       p4     p5      p6       p7      p8     p9      p10 */

P = {0.35 0.50 0.20 0.10   0.10   0.30    0.20 0.25 0.15 0.20,

     0.30 0.10 0.20 0.30   0.20   0.20    0.40 0.05 0.45    0.10,

     0.15 0.20    0.35 0.25   0.50   0.20    0.25 0.55 0.10 0.10,

     0.20 0.20 0.25 0.35   0.20   0.30    0.15 0.15 0.30    0.60};

/* expected values and variance for each ordinal variable */

Expected = Mean(P) // Var(P);

varNames = "X1":"X10";

print Expected[r={"Mean" "Var"} c=varNames];

/* test the RandMVOrd function */

Delta = {1  0.45 0.4   0.35 0.3    0.25 0.2    0.15 0.1    0.05,

        0.45 1 0.45 0.4    0.35 0.3    0.25 0.2    0.15 0.1,

        0.4  0.45 1   0.45    0.4    0.35 0.3    0.25 0.2    0.15,

        0.35  0.4 0.45 1    0.45 0.4    0.35 0.3    0.25 0.2,

        0.3  0.35 0.4   0.45     1    0.45 0.4    0.35 0.3    0.25,

        0.25  0.3 0.35 0.4    0.45   1    0.45 0.4    0.35 0.3,

        0.2  0.25 0.3   0.35    0.4    0.45 1    0.45 0.4    0.35,

        0.15  0.2 0.25 0.3   0.35 0.4    0.45   1    0.45 0.4,

        0.1  0.15 0.2   0.25   0.3    0.35 0.4    0.45   1    0.45,

        0.05  0.1 0.15 0.2 0.25 0.3    0.35 0.4   0.45     1};

call randseed(54321);

X = RandMVOrdinal(1000, P, Delta);

/* print results */

varNames = ('x1':'x10');

create MVO var {X1 X2 X3 X4 X5 X6 X7 X8 X9 X10};  append from var;  close MVO; /* write data */

run;              

[thanoon]

Hi

I am waiting anyone help me to correct the previous program please.

Regards

[Rick_SAS]

As I commented in another thread (https://communities.sas.com/thread/52019), it is extremely hard to invent means and covariances that are valid.  Try using smaller correlations. The following program works:

d = 1 || do(0.4, 0.04, -0.04);
Delta = toeplitz(d);

v = {0.05 0.05 0.05 0.05 0.10 0.10 0.10 0.10 0.2 0.20}`;
P = j(10,10);
P[,1] = v;
call randseed(1);
do i = 2 to 10;
   P[,i] = v[ranperm(10)];
end;
print P, Delta;
X = RandMVOrdinal(1000, P, Delta);

[thanoon]

thank you so much for your help i dont have any problem when i used small number of variables but in my research i need to simulate at least 10 variables so i forgot to put quotation marks in include statement after that i got on a results. thanks alot

[thanoon]

HI

when i simulating data i found some outliers but i dont write any condition to put outliers in data how can i solve this problem.

thanks in advance

[Rick_SAS]

If by "outlier" you mean "an observation from an area in which the probability density is low," then this isn't necessarily a problem. You expect to get a few points in those regions. For example, if you simulate 1 million observations from a univariate standard normal distribution, you should expect to get 63 observations such that |x|>4:

P = 2*1e6*Cdf("Normal", -4);

print P;

They are outliers, yes, but they are supposed to be there.  The same is true for other distributions.  It is only a "problem" if you are observing these low-probability regions more often than theory predicts.

[thanoon]

THANKS ALOT

[thanoon]

hi

if i want to simulate multi sample data can i just change the rand seed for each one.

thanks in advance

[Rick_SAS]

If you want to generate many samples in one program, you can use a loop or just generate one long matrix:

/* loop approach */

NumSamples = 10;

call randseed(54321);

do i = 1 to NumSamples;

   X = RandMVOrdinal(1000, P, Delta);

  /* do something with each sample  */

end;

/* or */

/* each block of 1000 is a sample */

X = RandMVOrdinal(NumSamples*1000, P, Delta);

[thanoon]

thanks dr. rick

i want to simulate different samples like different cultures or different countries.

thanks alot

[thanoon]

this code is ok now ?

proc iml;

load module=_all_;                     /* load the modules */

    /* P1    P2     P3       p4     p5      p6       p7      p8     p9      p10 */

P = {0.35 0.50 0.20 0.10   0.10   0.30    0.20 0.25 0.15 0.20,

     0.30 0.10 0.20 0.30   0.20   0.20    0.40 0.05 0.45    0.10,

     0.15 0.20    0.35 0.25   0.50   0.20    0.25 0.55 0.10 0.10,

     0.20 0.20 0.25 0.35   0.20   0.30    0.15 0.15 0.30    0.60};

/* expected values and variance for each ordinal variable */

Expected = Mean(P) // Var(P);

varNames = "X1":"X10";

print Expected[r={"Mean" "Var"} c=varNames];

/* test the RandMVOrd function */

Delta = {1  0.45 0.4   0.35 0.3    0.25 0.2    0.15 0.1    0.05,

        0.45 1 0.45 0.4    0.35 0.3    0.25 0.2    0.15 0.1,

        0.4  0.45 1   0.45    0.4    0.35 0.3    0.25 0.2    0.15,

        0.35  0.4 0.45 1    0.45 0.4    0.35 0.3    0.25 0.2,

        0.3  0.35 0.4   0.45     1    0.45 0.4    0.35 0.3    0.25,

        0.25  0.3 0.35 0.4    0.45   1    0.45 0.4    0.35 0.3,

        0.2  0.25 0.3   0.35    0.4    0.45 1    0.45 0.4    0.35,

        0.15  0.2 0.25 0.3   0.35 0.4    0.45   1    0.45 0.4,

        0.1  0.15 0.2   0.25   0.3    0.35 0.4    0.45   1    0.45,

        0.05  0.1 0.15 0.2 0.25 0.3    0.35 0.4   0.45     1};

/* loop approach */

NumSamples = 10;

call randseed(54321);

do i = 1 to NumSamples;

   X = RandMVOrdinal(1000, P, Delta);

  /* do something with each sample  */

/* print results */

c=varNames;

create MVO from X[colname=c];  append from X;  close MVO;

quit;

because i receive this warning

WARNING: DO group statements were not executed due to missing END at the time of QUIT.

[thanoon]

dear dr.

i still have some errors after putting the previous code can you correct it to me please

proc iml;

load module=_all_;                     /* load the modules */

    /* P1    P2     P3       p4     p5      p6       p7      p8     p9      p10 */

P = {0.35 0.50 0.20 0.10   0.10   0.30    0.20 0.25 0.15 0.20,

     0.30 0.10 0.20 0.30   0.20   0.20    0.40 0.05 0.45    0.10,

     0.15 0.20    0.35 0.25   0.50   0.20    0.25 0.55 0.10 0.10,

     0.20 0.20 0.25 0.35   0.20   0.30    0.15 0.15 0.30    0.60};

/* expected values and variance for each ordinal variable */

Expected = Mean(P) // Var(P);

varNames = "X1":"X10";

print Expected[r={"Mean" "Var"} c=varNames];

/* test the RandMVOrd function */

Delta = {1  0.45 0.4   0.35 0.3    0.25 0.2    0.15 0.1    0.05,

        0.45 1 0.45 0.4    0.35 0.3    0.25 0.2    0.15 0.1,

        0.4  0.45 1   0.45    0.4    0.35 0.3    0.25 0.2    0.15,

        0.35  0.4 0.45 1    0.45 0.4    0.35 0.3    0.25 0.2,

        0.3  0.35 0.4   0.45     1    0.45 0.4    0.35 0.3    0.25,

        0.25  0.3 0.35 0.4    0.45   1    0.45 0.4    0.35 0.3,

        0.2  0.25 0.3   0.35    0.4    0.45 1    0.45 0.4    0.35,

        0.15  0.2 0.25 0.3   0.35 0.4    0.45   1    0.45 0.4,

        0.1  0.15 0.2   0.25   0.3    0.35 0.4    0.45   1    0.45,

        0.05  0.1 0.15 0.2 0.25 0.3    0.35 0.4   0.45     1};

/* loop approach */

NumSamples = 2;

call randseed(54321);

do i = 1 to NumSamples;

   X = RandMVOrdinal(500, P, Delta);

  /* do something with each sample  */

   end;

/* print results */

c=varNames;

create MVO from X[colname=c];  append from X;  close MVO;

quit;

many thanks

[Rick_SAS]

Every time you exit SAS, you need to say

%include "RandMVOrd.sas";

again.

[thanoon]

thanks alot dr. wicklin

i got on result but the results are one sample only i need 2 samples (means two different data) and when i changed number of samples in sas program i got on 1 sample also  how can i simulate two different sample such n=500 , n1=250 , n2=250.

thanks again

[thanoon]

hi

i still wait anyone tell me how can i get on a two different data by using the above command.

regards